Table Of Content

Word-order and the Interpretation of Nominals in Plains Cree Jeff Mühlbauer 1. Introduction1 Until recently, Algonquian linguistic research has not considered word order and phrase structure to be particularly important components of these languages. The decision to exclude phrase structure from discussion is based on the seemingly erratic distribution of constituents in an Algonquian utterance when the traditional Algonquianist categories of noun, verb, and particle (i.e. everything that is not a noun or a verb) are the sole focus of study. This characterization is born out when a typical sentence is tested with native speakers (from Déchaine 1997). 1) a. wâpamê-w, nâpêw, êkoni atimw-a. VSO see-3, man, this dog-obv b. wâpamêw, êkoni atimwa, nâpêw. VOS c. nâpêw, wâpamêw êkoni atimwa. SVO d. êkoni atimwa, wâpamêw nâpêw. OVS e. nâpêw, êkoni atimwa, wâpamêw. SOV f. êkoni atimwa, nâpêw, wâpamêw. OSV The three Algonquianist categories of noun, verb, and particle (i.e. everything that is not a noun or a verb) indeed occur in any possible combination, and are at least superficially equivalent in the syntax. Word order does not appear to affect the assignment of grammatical functions. Thus, in a strictly predicational sense, Algonquian languages have "free" word order. However, there is strong evidence to suggest that these word orders are not equivalent in the semantics. In particular, the position of a nominal within the sentence is related to its interpretation, and has distinct consequences both for the semantics and the syntax. This can be demonstrated by tracking the position of nominals in large contextual spans, followed up with carefully-structured elicitation. Because this particular area of Algonquian linguistics is so poorly-studied, there is a great deal of work to be done. To that end, this paper is intended to be the outline of a research program in the semantic consequences of nominal position, with special focus on Plains Cree. 1 This work is supported in part through SSHRC grant (SSHRC 412-97-0016) to Anna-Marie DiSciullo and R.-M. Déchaine. Special thanks to Wally Awâsis, Clare Cook, Henry Davis, Rose-Marie Déchaine, Diana Gibraiel, Lisa Matthewson, and H.C. Wolfart. All mistakes are my own, and, as Al Gore once said of the Constitution, this paper is a living document. Key: 21=1pl Inclusive, OBV = obviative, ACC = nominative/accusative marker, ERG = ergative/absolutive marker, AS/IS =animate/inanimate subject, AO/IO=animate/inanimate object, conj=conjunct clause marker, REL=relative clause marker, NEG = negation, Q = question, STAT = stative, DIM = diminutive, POSS = possessor theme marker, USC = unspecified subject construction. JKN = Jim Kâ-Nîpitêhtêw Text. SW1 = Sarah Whitecalf's first lecture. 1 1.1 Linguist-produced Sentences vs. Native Speaker-Produced Sentences Many native speakers of Plains Cree have amazing flexibility in interpreting constituent structure (see §1). Given multiple constituents, many word orders can be made acceptable. Thus, if we are to begin a careful characterization of word order effects, a sharp line is needed between sentences produced by the linguist and those produced by the consultant. Linguists have used two different methods of testing sentence structure – texts and elicitation, with widely different results. This variation should not be surprising, since a text constitutes the uninterrupted discourse of a fluent speaker, whereas elicitation can involve linguists who do not necessarily understand the complete architecture of the language. Interestingly, freedom of word-order is not equally demonstrated by the linguist and the consultant. Although it is true that many word orders are deemed acceptable by speakers, the variety of word orders produced by the native speaker in a natural context shows a great deal less variation. Thus, for the purposes of this study, I will be preferential to sentences produced by native speakers without interference from linguists. Until we understand a great deal more about word order effects than we currently do, this can be the only reliable course of action. 1.1 Survey of the Field In his 1962 grammar of Menominee (an Algonquian language spoken in Wisconsin), Bloomfield summarizes Menominee's phrase structure in this way: The description of phrase structure is rendered partly indeterminate by stylistic variations of various sorts. The constituents of a phrase often appear in other than the usual order, or separated by other words, or with pause intonation between them, or with two or all three of these. (Bloomfield 1962:440) Wolfart does not discuss phrase structure at all in his 1973 grammar of Plains Cree, nor does Uhlenbeck in his 1938 grammar of Blackfoot. Thus some of the original major works on the western branch of Algonquian languages have nothing of use to say on the subject of word order. As Algonquian linguistics developed in the 1990s, several scholars began to seriously consider issues of word order. Ahenakew (1987) details the unique uses of the demonstrative ôma in Plains Cree, which shows strong ordering restrictions. Tomlin and Rhodes (1992) considered Ojibwa narratives, looking at the distribution of words, and determined that the neutral or "unmarked" order for Ojibwa was verb initial. Dahlstrom (1995) deals with topic and focus structure in Algonquian, determining that there is a topic position on the left edge of the verbal complex, followed by a modifier position. Finally, Wolfart (1996) comments on the intonational pauses in Plains Cree, and what they might mean for Plains Cree word order. Most recently, Déchaine et. al. (2003) has considered the ordering of modifiers, and Shields (2003) has looked at discourse roles of nominals and their ordering in Menominee. 2 In the broader world of linguistics, work on word order has been more prolific. Givón (1979) draws a distinction between word order possibilities in languages by characterizing a set of languages (e.g. Latin and Greek) that are "pragmatically- conditioned" in their word order. In his work on Walpiri, Hale coined the term "non- configurational" to characterize languages whose word-order is free. Baker (1996) and Jelinek (1984) have sought to encode this characterization in modern syntactic theory by positing that these languages introduce their arguments as morphology on the verb head, and thus all other DPs are adjuncts. In 1992, Hale refined his definition of "free" word order strictly to describe languages that do not have word order affect grammatical function. 1.2 Limits to Word Order Flexbility Despite the generalization repeated above, it is not entirely true that word order does not affect grammatical function. This is demonstrable in the work of several linguists, as well as my own fieldwork, and bears on the central concern of this paper – word-ordering of nominals – by showing that word-ordering touches many diverse areas of the grammar. Ahenakew (1987) and Déchaine (1997) consider the predicational use of demonstratives, which shows extremely strict word-order restrictions. 2) a. ôma maskisin b. maskisin ôma this shoe shoe this "this shoe" "This is a shoe" (Ahenakew 1987) Here, the position of the determiner with respect to the nominal has serious consequences for grammatical function. The prenominal demonstrative creates a DP, while the postnominal demonstrative creates a predicational structure. Further, this construction is productive in clefting contexts. 3) a. "êwak ôma kâ-mâcîhtâ-yân ôta," topic this REL-begin-1 here, "this which I began here" (JKN 46.2) b. "êkota ôma wîst ê-atoskêt ôta," there this 3-emph.PR CONJ-work-3 here "It is there that she also works," (JKN 46.29) Here, the clefting is used with a topic marker and a locative, and is crucial to maintenance and shifting of topics (see §4.2.1 below). This predication structure can even be used to cleft larger constituents, in what Ahenakew (1987) calls a "factitive" use of the demonstrative. 4) "kîkway anima k-âta-~-kakwê-kitâsômâyahk, êkâ kîkway ka-tôtahk anima," "kîkway anima k-âta-~-kakwê-kitâsô-m-â-yahk, êkâ kîkway ka-tôt-am-k anima something this chng-although-try.to-warn-by.mouth-ACC-21 NEG something REL-thus- 3 do-IO-3 this "it is that we have tried to warn them against something in vain, that they do not do something," (JKN 46.24-25) Here, the entire verb structure, along with negation and an indefinite pronoun, is in predicate position, with the distal demonstrative anima creating the predicational structure. These kinds of restrictions have also been discussed by Wolvengrey (2003). Some other data that bears on this issue comes from contexts that use overt nominals and adverbs. Given a verb, a noun, and an adverb, a native speaker can accept most word orders, with an interesting exception (from elicitation). 5) a. ni-npa-h-â-w ohcitaw amiskw 1-sleep-CAUS-ACC-3 on.purpose beaver "I killed a beaver on purpose" b. # ni-npa-h-â-w amisk ohcitaw 1-sleep-CAUSE-ACC-3 on.purpose beaver "I killed the beaver, who was a beaver on purpose." This second ordering can be rescued by right-dislocating the adverb with a pause. 6) ni-npa-h-â-w amiskw, ohcitaw 1-sleep-CAUSE-ACC-3 beaver, on.purpose "I killed a beaver on purpose." Thus word-ordering plays a crucial role in the resolution of adverbial modification. This finding is consistent with what Déchaine et. al. (2003) found in their study of modifier positioning. Other phenomena that show syntactic word-ordering restrictions include relative clauses, clefting constructions, reason clauses, and construct-state possession. An account of Plains Cree that assumes syntactically free word-ordering could not account for any these phenomena, since the grammatical functions crucially depend on the ordering of the constituents. Thus, despite the generalization that word order is usually "free" with respect to syntax, we should be aware that this, like many generalizatoins in Plains Cree, is not entirely true. 2 Macro-Syntax vs. Micro-Syntax Before going in to an account of the semantics and syntax of nominal positioning, it is important to consider what components of Plains Cree grammar are potentially accessible to operations that could impact word order, such as scrambling, extraction, left-dislocation, raising, etc. There are essentially two domains in Algonquian languages that are relevant to a syntactic and semantic analysis. For the purposes of this study, I will tentatively term these two domains "Macro-Syntax" and "Micro-Syntax". 4 2.1 Micro-Syntax The first domain to consider is Micro-Syntax. The ordering of constituents within this domain is generally rigid, allowing only specific local reorderings (see Cook 2003 on semantic and syntactic consequences of preverb ordering). In contrast to Macro-Syntax, these components are not able to be disconnected from one another. For example, when asking a question about an event in Plains Cree, the entire verbal complex must be fronted as a single constituent. 7) ki-kî-wâp-m-â-w cî ana nâpêw * kî cî kiwâpamâw ana nâpêw 2-perf-see-by.eye-ACC-3 Q that man kiwâpamâw cî ana nâpêw "Did you see that man?" Likewise, in relative clauses, extracting the nominal head does not affect the verb's pronominal marking. 8) ni-kî-wâp-m-â-w ana nâpêw kâ-ahko-si-t * ana nâpêw kâ-ahkosi_ 1-PERF-see-by.eye-ACC-3 that man REL-sick-STAT-3 nkîwâpmâw ana nâpêw kâhkosit "I saw the man who was sick." This is in sharp contrast to languages like English, which allow the extraction of a single functional component (e.g. tense with do-support). The domain of Micro-Syntax encompasses all head-marking nominal and verbal morphology. For nominals, this includes pronominal marking, plural and obviation marking, non-verbal adjectives, and the possessor marker, as shown in 9. 9) [PERS [PREN1 [PREN2 [NOUN] IM] DIM]STEM PERSNUM] NUM]2 ki-wâpiski-pôsîs-im-sîs-inaw-ak 2-white-cat-poss-dim-21pl-pl kiwâpskipôsimsisnawak "our white kittens" For verbs, Micro-Syntax includes impersonal marking, a subset of aspectual and modal markers, transitivity suffixes, incorporated nouns, ergative/accusative marking, and so on.3 This is all in addition to the pronominal marking seen on nouns, making for a syntactically complex structure. 10) [PERS [PREV* [ROOT] MANNER] ERG/ACC] PERS] NUM] STEM ni-kî-wâp-m-â-w-ak 1-perf-see-by.eye-acc-3-pl 2 For the reasoning behind 2 prenoun positions, see Mühlbauer (2003b). 3 See Wolfart (1973) for a full treatment of all nominal and verbal morphology. 5 nkîwâpmâwak "I have seen them." Several linguists have attempted to account for the structure of these Micro- Syntactic components, including Déchaine and Reinholz (1997), Déchaine (1999) and Hirose (2000). In these accounts, the focus has been on the specific syntactic functions that these morphemes have. For example, Hirose (2000) shows that there is a one-to-one correspondence between syntactic arguments and stem-internal verbal suffixes. Therefore these morphemes correlate with transitivity marking, which translates to a vP shell in a syntactic analysis like Kratzer (1993). 11) wâpa-m-ê-w "he sees him" vP V pro vP V v vP -m- V pro vP V v -ê- Déchaine (1999) deals with the question of linear-ordering of these elements by analyzing it as a case of mixed-affixation. A noun, for example, moves through a series of spec positions and checks features off. These higher heads attach to the noun via phrasal encliticization, either as prefixes or suffixes. 12) DP V D PERSP V PERS NUMP V NUM NP V N Thus, in these analyses, Micro-Syntax reduces to phonologically-dependent morphemes that carry simplex syntactic content. Note that some of the syntactic structure spelled-out in Plains Cree is also apparent in English, although it is morphologically covert. A verb such as "kick" is encoded for a certain transitivity and telicity, but is a single morphological component 6 where Plains Cree would have a minimum of three. These covert pieces of "kick" can be manipulated via modification (e.g. "John was kicked" or "John kicks little children"), but they cannot be reordered or extracted. Thus the difference between a language like English and a language like Plains Cree reduces, in some respects, to an issue of morphological span (c.f. Williams, Déchaine 2003). This means that the existence of Micro-Syntax is not an Algonquian-specific phenomenon, which would be apparent to anyone who has tried to determine the meaning of English words like "he", "the", "-ed", or even "kick" without presenting a speaker with a full, well-formed utterance. As for the difference in morphological span between Algonquian languages and other human languages, it is my opinion that this is related to the phonological structure of these languages (e.g. syllable structure, phonemic inventory, etc). Whether it is the phonology that causes the syntax or the other way around is well beyond my current understanding. 2.2 Macro-Syntax Macro-Syntax operates on the output of Micro-Syntax, such that it manipulates syntactically-complex constituents as units. Thus, in a DP like Wâpastim o-maskisin-a "Wâpastim his-shoe-s", Wâpastim and omaskisina can be moved or extracted, but internal components of either (i.e. Micro-Syntax) can not be. Since this paper is concerned with the word order effects of nominals, it is Macro-Syntax that will be central. If we assume that Plains Cree has a neutral word order for constituents, any deviation from this neutral order is an instance of Macro-Syntactic operations. This likely includes question formation (13a), topicalization (13b), dislocation (13c), and rhetorical strategies like chiasma (13d). 13) a. kikîwâpamâw cî ana nâpêw 2-perf-see-by.eye-AO-3 ques that man "Did you see that man?" b. "…kiyânaw tipiyaw kitawâsimisinaw kôsisiminaw, kiyâm pikoyikohk ka- kisîwêhkahtawâyahk kîkway ê-pakwâtamawâyahk, …" 21 personally 2-child-1pl 2-grandchild-1pl, let.it.be no.matter future- speak.loudly.at-AO-21 something conj-disapprove.TA-applic-dir-21, "….our own children and grandchildren, let it be that we speak loudly at them to disapprove of them for something, …" (JKN 46.21-22) c. "…tânisi kîkway k-êsi-nahêyihtahk, otawâsimisimâw oyôsisimimâw, …" how.IPC something.PR chng-thus.RR-have.peace.of.mind.TI-TI-3s, 3-child-poss-dim- imp-dir-3s 3-grandchild-poss-imp-dir-3s, "…how they would have some peace of mind, children and grandchildren" (JKN 46.19-20) 7 d. "…tahk âwiyâk kîkway nikî-ati-miyikwak ê-miywâsik, …" more.and.more someone something 1-perf-proceed-give-ERG-3-3pl CONJ-good-STAT-3 "and then they gave me more and more of that which is good." (JKN 48.6) Thus the crucial thing to determine is what the neutral word order is, so that we know what a deviation is, and therefore what must be accounted for as a Macro-Syntactic operation. To do this, I will begin by looking at modifier positioning. This will give us some structure to begin measuring the position of nominals. 3 Modifier Positions Déchaine, et. al (2003) analyzed a speech by the Plains Cree elder Kâ- pimwêwêhahk, paying particular attention to the ordering of modifying adjuncts, such as locative markers like êkota "there", adverbs like wêtinahk "quietly", and temporal markers like anohc "today". When all occurrences of modifiers were assembled, the result was a surprisingly structured hierarchy. 14) a. Preverbal b. Postverbal [TEMP > ADV* > LOC > TEMP > V] [V > ADV/LOC > TEMP] TEMP TEMP V V TEMP ADV* ADV/LOC TEMP V V ADV LOC VP ADV/LOC V LOC TEMP V TEMP VP Preverbal modifiers are more common, and are more often stacked. Their order is consistent throughout the text considered. Postverbal modifiers, on the other hand, are much less common, show limited ordering, and do not allow stacking. This would be consistent with a verb-fronting analysis, in which the verb raises to some position higher than one of the adverbial positions. Predicate modifiers occur internal to topic markers like clefts (c.f. Déchaine 1997) and other clause-typing elements like negation (c.f. Déchaine and Wiltschko 2002). This is consistent with Cinque's (1999) hierarchy of functional heads, and provides an important framework to begin considering where nominals fit in Plains Cree word ordering. 4 Nominal Positions Syntactically, the core component of an utterance is the verb. With the exception of structural cases (Bittner and Hale (1996)), the verb and its functional heads license the 8 existence of DPs and other elements. Semantically, the verb provides the core content of the sentence, taking an individual and forming a function of individuals to truth conditions. This centrality gives us a good foothold on the word order of Plains Cree. If we look at nominals through the lens of their relation to the verb, we can begin to make generalizations that are reasonably grounded in the semantics and syntax. 4.2 The Preverbal Domain Preverbal nominals are less common than their postverbal counterparts, and bear specific functions. This, combined with the previous work done on the ordering patterns of predicate modifiers, provides us with good motivation to choose preverbal nominals as the place to begin this description. To this end, I will begin with the leftmost positions in the preverbal domain and move inwards. 4.2.1 Topic The notion of "Topic" that I will use for this analysis is tentatively defined as "what the discourse is about". Since topic is a difficult thing to get one's hands on in any concrete way, I will make heavy use of a counselling speech by the Plains Cree elder Kâ- pimwêwêhahk that I have already carefuly dissected for its rhetorical import (c.f. Déchaine et. al. 2003). This will help me in determining what the discourse is "about". With this in mind, Plains Cree seems to have what I will call an "inner" and an "outer" topic. These are differentiated by their position relative to the rest of the clause. 4.2.1.1 Outer Topic Outer Topics are characterized by left-dislocation effects similar in semantic and prosodic content to English's left-dislocation phenomenon. Typically, outer topics are separated from the main clause with an intonation or pause, and refer to a familiar discourse referent. 15) a. "…kiyânaw tipiyaw kitawâsimisinaw kôsisiminaw, kiyâm pikoyikohk ka- kisîwêhkahtawâyahk kîkway ê-pakwâtamawâyahk, …" 21 personally 2-child-1pl 2-grandchild-1pl, let.it.be no.matter future- speak.loudly.at-AO-21 something CONJ-disapprove.TA-applic-dir-21, "….our own children and grandchildren, let it be that we speak loudly at them to disapprove of them for something, …" (JKN 46.21-22) b. "êkosi aniki aniki môniyâsak wîstawâw, 'kiyâm' nika-itwân, 'tita- kitimâkinâkoyahkok ôma ê-nêhiyâwiyahk,'" so those those white.man-pl 3.emph.pn, let.it.be 1-fut-say-local, future-respect- ERG-21-3 this conj-cree-21, "And for the other Whites, too, 'Let them respect,' I will say, 'that we are Cree,'" (SW1) 9 c. " - ahpô ayisk môniyâsak, wîstawâw ê-tâpwêwakêyihtahkik êtikwê anima nipâkwêsimowin," or even white.man-PL, 3.emph.pn conj-true-by.mind-IO-3-3PL apparently that sun.dance-NOM," "- for even the Whites themselves must believe in the Sundance," (SW1) Here, the topic is separated off from the main clause with a pause (marked by Wolfart as a comma). The use of left dislocation in all of these examples coincides with a shift in the topic under discussion. Specifically, the predicate remains notionally constant, but what it is about (i.e. the topic) has changed. This is a kind of rhetorical parallelism, used to build a connection between ideas. In 15a, it is a shift from the traditional times of grandchildren being instructed by the proto-typical old woman to the obstinate grandchildren of those in the room. In 15b, it is a shift from the specific example of H.C. Wolfart's respect for Cree ways, to white men in general. In 15c, it is a shift from the Plains Cree's faith in the Sun Dance to the likelihood of the Whites faith in it. In all these instances, this new topic had already been under discussion previously, and thus left dislocation signals a return to an older topic. 4.2.1.2 Inner Topic Topicalized nominals may also occur as the left-most constituent inside the clause, preceding clause-typing components like namoya and êka "negation" (c.f. Déchaine and Wiltschko 2002 for the clausal status of negation), and subjunctive-type markers like kiyâm "let it be". 16) a. "êwakw âwa ayamihewiyiniw kâ-wî-petât," resump this pray-person REL-intend-bring-ACC-3 "that priest is going to bring it," (JKN 50.28) b. "otôsk-âyima êkâ kwayask ê-isi-wîcêhtoyit." 3-young-person-POSS-OBV NEG properly CONJ-thus-get.along-RECIP-3 "If their young people do not get along with one another." (JKN 48.28) c. "êwakw ânima pêyak kisêyiniw ê-kî-nakiskawak," topic that one old-man conj-perf -meet-by.foot-1>3 "I met a certain old man about that," (JKN 46.6) Here, the topic is not preposed and separated by a pause, as with Outer Topics. Inner topic is often used to introduce new discourse referents, who will be the subject of some span of discourse. Once this introductory topic-marking has been made, the nominals can be returned to a post-verbal position, as is shown in the beginning of a sacred story (âtayôhkêwin) told to Bloomfield (1930) by Adam Sâkêwêw. 10